- Email: [email protected]
Cancer survival statistics should be viewed with caution
Reflection and Reaction
figure should be kept in mind by health-care authorities, especially in the developing world, when compiling their list of disease priorities. Another aspect to be considered is the price of cancer drugs. The formula by which pharmaceutical industries calculate the costs of drugs is not transparent. With the introduction of generic medications in Brazil, the price of certain anticancer drugs has been reduced by almost 90%. This discrepancy in the pricing of identical trade and generic drugs suggests that manufacturers are setting prices in a way that allows a large profit margin to recoup their expenses handsomely. For example, in Brazil, the average cost of one milligram of pegylated interferon is about US$2·60 compared with one milligram of gold that was recently trading at about US$0·026! A further important point is that nowadays, patients and their relatives have considerable access to medical information. Although such access has obvious advantages, it has also added a new and unpredictable component to government health-care budgets. Because cancer is universally seen as a fatal disease, legal demands from patients often force health ministries to purchase new, supposedly efficacious, drugs even when their cost–benefit is not yet known.
Since 2005, about 10 000 legal demands for purchasing new drugs have been successful in southern Brazil. In general, judges are being misled into making decisions on the basis of medical recommendations that have not been based on published consensus papers, findings of meta-analyses, or therapeutic guidelines produced by WHO or other international medical organisations. We will certainly use the information provided by Rawlins to guide our health-care authorities in dealing with these important issues. In Brazil, the average per-capita annual gross domestic product is slightly over US$4000 and few, if any, of the new cancer drugs would fall within the WHO cost-effectiveness bracket in our country.4 *Gilberto Schwartsmann, Paulo D Picon Faculty of Medicine, Federal University of Rio Grande do Sul, Porto Alegre, Brazil [email protected] The authors declared no conflicts of interest. 1 2 3
4
Rawlings M. Paying for modern cancer care—a global perspective. Lancet Oncol 2007; 8: 749–51. Coates PM. Evidence-based reviews in support of health policy decisions. J Natl Cancer Inst 2007; 99: 1059. Stenberg K, Johns B, Scherpbier RW, Edejer TT. A financial road map to scaling up essential child health interventions in 75 countries. Bull World Health Organ 2007; 85: 305–14. Edejer TT, Baltussen R, Adam T, et al (eds). WHO guide to cost-effectiveness analysis. Geneva: World Health Organization, 2003.
Cancer survival statistics should be viewed with caution In the EUROCARE-4 study, recently published in The Lancet Oncology, Berrino and co-workers report the 5-year relative survival for eight cancers in patients from 23 countries.1 Relative survival is deemed to be an indicator of the quality of disease management and access to efficient treatment. Other factors that might affect survival estimates, such as lead-time bias (where survival is apparently longer because of diagnosis before the cancer is clinically evident) and incomplete registration of incidence, were discussed. However, Berrino and co-workers did not mention lengthtime bias—ie, the greater chance of finding indolent cancers when screening methods are used.2 Many indolent cancers would probably never be life threatening and, thus, would never affect mortality. Moreover, if no earlydetection method was used, many patients diagnosed with indolent tumours would not have been included in the cohort of patients for which 5-year survival was calculated. 1050
Cancers in the EUROCARE-4 study that were screendetectable between 1995 and 1999 were prostate cancer (by serum-prostate-specific antigen [PSA] testing), breast cancer (by mammography), and cutaneous melanoma (by increasing excision of pigmented lesions because of greater melanoma awareness by patients and by doctors). For the five other cancers (lung, colorectal, ovarian, and testicular cancer, and Hodgkin’s lymphoma), screening was less common (colorectal cancer) or no screening test existed in the period between 1995 and 1999. The basic formula for calculation of 5-year survival is (CX-d)/CX, where CX is the number of patients diagnosed with a specific cancer (X) and d is the number of these patients who died from any cause in the 5 years after diagnosis (the reasoning is similar for the 5-year relative survival used in the EUROCARE-4 study). If mortality from cancer X decreases, then d will decrease and survival will increase. But increased detection of indolent cancers http://oncology.thelancet.com Vol 8 December 2007
Reflection and Reaction
will increase CX without changing d—ie, the incidence and survival will increase, but mortality will not decrease, and survival will become steadily more associated with incidence and steadily less associated with mortality. To show the importance of length-time bias, we compiled cancer incidence and mortality data from 1995 to 2000 for each country included in the EUROCARE-4 study. We left out three countries because of small populations (Iceland and Malta) or because of absence of incidence data (Belgium). Incidence data by sex were extracted from the database Cancer Incidence in Five Continents3 and other publicly available databases. Similar data for mortality were extracted from the WHO mortality database.4 Incidence and mortality rates were ageadjusted using the World Standard Population. To analyse the relations between variables, we calculated Pearson’s correlation coefficients (r). The data in table 1 show that countries with high 5-year relative survival for prostate cancer also had a high incidence of this cancer, whereas 5-year relative survival was not associated with prostate-cancer mortality. By use of the formulae above, a 5-year survival of 54% (as found in the Czech Republic) could be equal to 460 deaths occurring during the 5 years after a diagnosis of prostate cancer in 1000 men. The ratio of incidences in Austria and in the Czech Republic in 1995 was 2·2 (ie, 68·4/31·0). If the number of men diagnosed with prostate cancer increased by a factor 2·2 without changes in mortality, the 5-year relative survival would jump to 79%, which is close to the 5-year relative survival recorded for Austria. A survey done in the European Union in 20065 reported the percentages of men aged 15 years or more who had undergone a PSA test in the previous 12 months. In view of our argument, we assumed that the prevalence of screening in this survey was similar to the prevalence of screening that prevailed at the end of the 1990s. The proportions of men who had undergone a PSA test was positively correlated with 5-year relative survival (r=0·76, p=0·0029) and with prostate-cancer incidence in 2000 (r=0·69, p=0·010). Of the 11 countries with total national expenditure for health above US$1999 per year (table 1),1 in Italy, the prostate cancer incidence (38·4 per 100 000 in 1995) and mortality (10·4 per 100 000 in 2000) were low relative to other countries. The mean prostate-cancer incidence in 1995 and mean mortality in 2000 were 35·3 and 17·7 per 100 000 respectively, for Denmark and the UK combined. http://oncology.thelancet.com Vol 8 December 2007
For the eight other countries, the mean incidence in 1995 was 59·3 and and the mean mortality in 2000 was 17·3. These substantial differences in incidence, but similarities in mortality, might explain why the 5-year relative survival was worse in Denmark and the UK despite these countries having comparable total national expenditure for health with the other countries. In this respect, an increased incidence of indolent cancers due to differential use of screening between countries is likely to give the impression of an association between cancer survival and economic resources devoted to health care. Table 2 shows that in the EUROCARE-4 study, the highest variability in 5-year relative survival was for cancers subject to screening (prostate and breast cancer, and cutaneous melanoma), and the lowest variability was for cancers for which screening methods were not available (ovarian and lung cancer). The 5-year relative survival of screendetectable cancers was positively correlated with incidence, but had no correlation with mortality. For cancers for which no screening existed (ovarian and lung cancer), a significant negative correlation between 5-year relative survival and mortality was noted. The 5-year relative survival was also negatively correlated with incidence Country
5-year relative survival, %1
Incidence per 100 000
Mortality per 100 000
Men who underwent a PSA test in past 12 months, %5
1995–1999
1995
2000
1995
2000
Austria*
84·9
68·4
89·5
17·5
16·3
28
Switzerland*
82·9
58·5
86·8†
19·2
17·5
NA
Germany*
81·6
47·3
54·2‡
17·0
14·4
21
Netherlands*
80·9
56·1
56·6
19·6
17·5
9
Finland*
80·0
62·4
82·3
18·0
17·3
14
Italy*
79·6
38·4
44·4‡
11·2
10·4
15
France*
79·1
57·5
75·3
15·5
14·9
15
Sweden*
77·3
62·7
86·2
21·7
21·9
13
Spain
74·7
36·0
39·6‡
13·9
12·4
10 NA
Norway*
74·5
61·1
81·8
22·8
23·2
UK England*
69·8
39·7
51·2
17·1§
14·8§
UK Wales*
68·7
43·1
55·5
··
··
UK Scotland*
67·8
45·1
45·4
17·0
15·8
UK Northern Ireland*
60·8
40·0
45·4
16·2
15·2
··
Poland
60·5
17·6
21·7‡
11·1
12·6
6
Slovenia
58·2
27·9
33·3
14·8
17·7
8
Czech Republic
54·4
31·0
35·5
15·9
16·8
6
Denmark*
47·7
28·7
39·7
20·1
20·1
5
8¶ ·· ··
Countries are ranked according to decreasing 5-year relative survival. *Countries with Total National Expenditures on Health ≥US$2000 per year. †Last year available was 1999 . ‡Last year available was 1997. §Data for England and Wales combined. ¶Data for all countries in UK combined. NA=not available.
Table 1: Incidence and mortality from prostate cancer in 15 countries included in the EUROCARE-4 study
1051
Reflection and Reaction
Difference between highest and lowest 5-year relative survival between 1995 and 1999,%
Correlation (r)† between 5-year relative survival in the period 1995–1999 and incidence in 1995
Correlation (r)† between 5-year relative survival in the period 1995–1999 and mortality in 2000
Prostate
37·2
0·79‡
Breast
15·0
0·54§
–0·11 –0·15
Cutaneous melanoma
29·1
0·42
–0·03
Colorectal
20·9
0·14
–0·58§
Ovary
12·5
–0·48
–0·63§
Lung
6·4
–0·43
–0·57§
Testis cancer and Hodgkin’s lymphoma not included because of low mortality. †Pearson’s correlation coefficient derived from least squares linear regression. ‡p≤0·001. §0·05>p>0·001.
Table 2: Relation between 5-year relative survival, incidence, and mortality in countries included in the EUROCARE-4 study
because of the rapid fatal outcome usually associated with these two cancers. The need to interpret survival data by considering length-time bias is supported by findings that the steep increases in the incidence of prostate cancer, breast cancer, and cutaneous melanoma, recorded in most European countries, essentially concern early-stage cancers,6–8 many of which would probably never be life-threatening during the 5-year period after diagnosis. After introduction of a screening test, the lead-time bias can increase the incidence by a maximum of one multiplied by the mortality rate. Any further increases in incidence might be caused by length-time bias. In table 1, length-time bias might have started to affect 5-year relative survival in the UK, and Austria might be the country where its effect was the highest. In addition to length-time bias, the EUROCARE-4 study only included 1–58% of the population in nine of the countries studied—this has unknown effects on survival estimates. For example, most Italian data came from the north of the country,1 where the incidence of and survival from prostate cancer are higher than in the south.9 Relative survival data are susceptible to biases and limitations of both incidence and mortality indicators.10 Mortality data are usually available for entire countries and not confined to some sub-country areas. Therefore, in the current situation, mortality data remain the most reliable indicator for the international comparison of outcomes in cancer. *Philippe Autier, Mathieu Boniol, Clarisse Héry, Eric Masuyer, Jacques Ferlay International Agency for Research on Cancer, 69372 Lyon Cedex 08, France [email protected] The authors declared no conflicts of interest.
1052
1
2 3 4 5 6
7
8
9
10
Berrino F, De Angelis R, Sant M, et al. Survival for eight major cancers and all cancers combined for European adults diagnosed in 1995–99: results of the EUROCARE-4 study. Lancet Oncol 2007; 8: 773–83. Sackett DL, Haynes RB, Guyatt GH, Tugwell P. Clinical epidemiology: a basic science for clinical medicine, 2nd edn. Boston: Little, Brown, 1991: 153–70. Parkin DM, Whelan SL, Ferlay J, Storm H. Cancer Incidence in Five Continents, Volume I to VIII. Lyon: IARC CancerBase No.7, 2005. World Health Organisation Databank, Geneva, Switzerland. WHO Statistical Information System. http:www-dep.iarc.fr (accessed Nov 9, 2007). Health in the European Union, Report of September 2007. http://ec.europa. eu/health/ph_publication/eb_health_en.pdf (accessed Nov 9, 2007). Fracheboud J, Otto SJ, van Dijk JA, et al. Decreased rates of advanced breast cancer due to mammography screening in the Netherlands. Br J Cancer 2004; 91: 861–67. Helgesen F, Holmberg L, Johansson JE, Bergström R, Adami HO. Trends in prostate cancer survival in Sweden, 1960 through 1988: evidence of increasing diagnosis of nonlethal tumors. J Natl Cancer Inst 1996; 88: 1216–21. Lipsker DM, Hedelin G, Heid E, Grosshans EM, Cribier BJ. Striking increase of thin melanomas contrasts with stable incidence of thick melanomas. Arch Dermatol 1999; 135: 1451–56. Quaglia A, Vercelli M, Puppo A, et al. Prostate cancer in Italy before and during the ‘PSA era’: survival trend and prognostic determinants. Eur J Cancer Prev 2003; 12: 145–52. Doll R, Peto R. The causes of cancer. Appendix C. Oxford: Oxford University Press, 1981.
In a recent issue of The Lancet Oncology, findings from EUROCARE-41 were published, which showed that survival for all cancers combined in the UK was not only below the European average, but also similar to survival in some eastern European countries, where the healthcare budget is less than a third of that in the UK. These findings led to the suggestion that the UK government now faces substantial challenges in improving cancer survival.2 The study also showed that European survival from solid tumours was lower than that reported in the USA.3 The clear implication of these findings is that patients presenting with the same disease will do less well in the UK than in other western countries, a thought that is frightening for UK National Health Service (NHS) patients and demoralising for NHS staff. However, a closer look at the published data suggests that the quality of care offered by the NHS is not the only explanation for poor UK survival, with differences in patient characteristics, disease http://oncology.thelancet.com Vol 8 December 2007
figure should be kept in mind by health-care authorities, especially in the developing world, when compiling their list of disease priorities. Another aspect to be considered is the price of cancer drugs. The formula by which pharmaceutical industries calculate the costs of drugs is not transparent. With the introduction of generic medications in Brazil, the price of certain anticancer drugs has been reduced by almost 90%. This discrepancy in the pricing of identical trade and generic drugs suggests that manufacturers are setting prices in a way that allows a large profit margin to recoup their expenses handsomely. For example, in Brazil, the average cost of one milligram of pegylated interferon is about US$2·60 compared with one milligram of gold that was recently trading at about US$0·026! A further important point is that nowadays, patients and their relatives have considerable access to medical information. Although such access has obvious advantages, it has also added a new and unpredictable component to government health-care budgets. Because cancer is universally seen as a fatal disease, legal demands from patients often force health ministries to purchase new, supposedly efficacious, drugs even when their cost–benefit is not yet known.
Since 2005, about 10 000 legal demands for purchasing new drugs have been successful in southern Brazil. In general, judges are being misled into making decisions on the basis of medical recommendations that have not been based on published consensus papers, findings of meta-analyses, or therapeutic guidelines produced by WHO or other international medical organisations. We will certainly use the information provided by Rawlins to guide our health-care authorities in dealing with these important issues. In Brazil, the average per-capita annual gross domestic product is slightly over US$4000 and few, if any, of the new cancer drugs would fall within the WHO cost-effectiveness bracket in our country.4 *Gilberto Schwartsmann, Paulo D Picon Faculty of Medicine, Federal University of Rio Grande do Sul, Porto Alegre, Brazil [email protected] The authors declared no conflicts of interest. 1 2 3
4
Rawlings M. Paying for modern cancer care—a global perspective. Lancet Oncol 2007; 8: 749–51. Coates PM. Evidence-based reviews in support of health policy decisions. J Natl Cancer Inst 2007; 99: 1059. Stenberg K, Johns B, Scherpbier RW, Edejer TT. A financial road map to scaling up essential child health interventions in 75 countries. Bull World Health Organ 2007; 85: 305–14. Edejer TT, Baltussen R, Adam T, et al (eds). WHO guide to cost-effectiveness analysis. Geneva: World Health Organization, 2003.
Cancer survival statistics should be viewed with caution In the EUROCARE-4 study, recently published in The Lancet Oncology, Berrino and co-workers report the 5-year relative survival for eight cancers in patients from 23 countries.1 Relative survival is deemed to be an indicator of the quality of disease management and access to efficient treatment. Other factors that might affect survival estimates, such as lead-time bias (where survival is apparently longer because of diagnosis before the cancer is clinically evident) and incomplete registration of incidence, were discussed. However, Berrino and co-workers did not mention lengthtime bias—ie, the greater chance of finding indolent cancers when screening methods are used.2 Many indolent cancers would probably never be life threatening and, thus, would never affect mortality. Moreover, if no earlydetection method was used, many patients diagnosed with indolent tumours would not have been included in the cohort of patients for which 5-year survival was calculated. 1050
Cancers in the EUROCARE-4 study that were screendetectable between 1995 and 1999 were prostate cancer (by serum-prostate-specific antigen [PSA] testing), breast cancer (by mammography), and cutaneous melanoma (by increasing excision of pigmented lesions because of greater melanoma awareness by patients and by doctors). For the five other cancers (lung, colorectal, ovarian, and testicular cancer, and Hodgkin’s lymphoma), screening was less common (colorectal cancer) or no screening test existed in the period between 1995 and 1999. The basic formula for calculation of 5-year survival is (CX-d)/CX, where CX is the number of patients diagnosed with a specific cancer (X) and d is the number of these patients who died from any cause in the 5 years after diagnosis (the reasoning is similar for the 5-year relative survival used in the EUROCARE-4 study). If mortality from cancer X decreases, then d will decrease and survival will increase. But increased detection of indolent cancers http://oncology.thelancet.com Vol 8 December 2007
Reflection and Reaction
will increase CX without changing d—ie, the incidence and survival will increase, but mortality will not decrease, and survival will become steadily more associated with incidence and steadily less associated with mortality. To show the importance of length-time bias, we compiled cancer incidence and mortality data from 1995 to 2000 for each country included in the EUROCARE-4 study. We left out three countries because of small populations (Iceland and Malta) or because of absence of incidence data (Belgium). Incidence data by sex were extracted from the database Cancer Incidence in Five Continents3 and other publicly available databases. Similar data for mortality were extracted from the WHO mortality database.4 Incidence and mortality rates were ageadjusted using the World Standard Population. To analyse the relations between variables, we calculated Pearson’s correlation coefficients (r). The data in table 1 show that countries with high 5-year relative survival for prostate cancer also had a high incidence of this cancer, whereas 5-year relative survival was not associated with prostate-cancer mortality. By use of the formulae above, a 5-year survival of 54% (as found in the Czech Republic) could be equal to 460 deaths occurring during the 5 years after a diagnosis of prostate cancer in 1000 men. The ratio of incidences in Austria and in the Czech Republic in 1995 was 2·2 (ie, 68·4/31·0). If the number of men diagnosed with prostate cancer increased by a factor 2·2 without changes in mortality, the 5-year relative survival would jump to 79%, which is close to the 5-year relative survival recorded for Austria. A survey done in the European Union in 20065 reported the percentages of men aged 15 years or more who had undergone a PSA test in the previous 12 months. In view of our argument, we assumed that the prevalence of screening in this survey was similar to the prevalence of screening that prevailed at the end of the 1990s. The proportions of men who had undergone a PSA test was positively correlated with 5-year relative survival (r=0·76, p=0·0029) and with prostate-cancer incidence in 2000 (r=0·69, p=0·010). Of the 11 countries with total national expenditure for health above US$1999 per year (table 1),1 in Italy, the prostate cancer incidence (38·4 per 100 000 in 1995) and mortality (10·4 per 100 000 in 2000) were low relative to other countries. The mean prostate-cancer incidence in 1995 and mean mortality in 2000 were 35·3 and 17·7 per 100 000 respectively, for Denmark and the UK combined. http://oncology.thelancet.com Vol 8 December 2007
For the eight other countries, the mean incidence in 1995 was 59·3 and and the mean mortality in 2000 was 17·3. These substantial differences in incidence, but similarities in mortality, might explain why the 5-year relative survival was worse in Denmark and the UK despite these countries having comparable total national expenditure for health with the other countries. In this respect, an increased incidence of indolent cancers due to differential use of screening between countries is likely to give the impression of an association between cancer survival and economic resources devoted to health care. Table 2 shows that in the EUROCARE-4 study, the highest variability in 5-year relative survival was for cancers subject to screening (prostate and breast cancer, and cutaneous melanoma), and the lowest variability was for cancers for which screening methods were not available (ovarian and lung cancer). The 5-year relative survival of screendetectable cancers was positively correlated with incidence, but had no correlation with mortality. For cancers for which no screening existed (ovarian and lung cancer), a significant negative correlation between 5-year relative survival and mortality was noted. The 5-year relative survival was also negatively correlated with incidence Country
5-year relative survival, %1
Incidence per 100 000
Mortality per 100 000
Men who underwent a PSA test in past 12 months, %5
1995–1999
1995
2000
1995
2000
Austria*
84·9
68·4
89·5
17·5
16·3
28
Switzerland*
82·9
58·5
86·8†
19·2
17·5
NA
Germany*
81·6
47·3
54·2‡
17·0
14·4
21
Netherlands*
80·9
56·1
56·6
19·6
17·5
9
Finland*
80·0
62·4
82·3
18·0
17·3
14
Italy*
79·6
38·4
44·4‡
11·2
10·4
15
France*
79·1
57·5
75·3
15·5
14·9
15
Sweden*
77·3
62·7
86·2
21·7
21·9
13
Spain
74·7
36·0
39·6‡
13·9
12·4
10 NA
Norway*
74·5
61·1
81·8
22·8
23·2
UK England*
69·8
39·7
51·2
17·1§
14·8§
UK Wales*
68·7
43·1
55·5
··
··
UK Scotland*
67·8
45·1
45·4
17·0
15·8
UK Northern Ireland*
60·8
40·0
45·4
16·2
15·2
··
Poland
60·5
17·6
21·7‡
11·1
12·6
6
Slovenia
58·2
27·9
33·3
14·8
17·7
8
Czech Republic
54·4
31·0
35·5
15·9
16·8
6
Denmark*
47·7
28·7
39·7
20·1
20·1
5
8¶ ·· ··
Countries are ranked according to decreasing 5-year relative survival. *Countries with Total National Expenditures on Health ≥US$2000 per year. †Last year available was 1999 . ‡Last year available was 1997. §Data for England and Wales combined. ¶Data for all countries in UK combined. NA=not available.
Table 1: Incidence and mortality from prostate cancer in 15 countries included in the EUROCARE-4 study
1051
Reflection and Reaction
Difference between highest and lowest 5-year relative survival between 1995 and 1999,%
Correlation (r)† between 5-year relative survival in the period 1995–1999 and incidence in 1995
Correlation (r)† between 5-year relative survival in the period 1995–1999 and mortality in 2000
Prostate
37·2
0·79‡
Breast
15·0
0·54§
–0·11 –0·15
Cutaneous melanoma
29·1
0·42
–0·03
Colorectal
20·9
0·14
–0·58§
Ovary
12·5
–0·48
–0·63§
Lung
6·4
–0·43
–0·57§
Testis cancer and Hodgkin’s lymphoma not included because of low mortality. †Pearson’s correlation coefficient derived from least squares linear regression. ‡p≤0·001. §0·05>p>0·001.
Table 2: Relation between 5-year relative survival, incidence, and mortality in countries included in the EUROCARE-4 study
because of the rapid fatal outcome usually associated with these two cancers. The need to interpret survival data by considering length-time bias is supported by findings that the steep increases in the incidence of prostate cancer, breast cancer, and cutaneous melanoma, recorded in most European countries, essentially concern early-stage cancers,6–8 many of which would probably never be life-threatening during the 5-year period after diagnosis. After introduction of a screening test, the lead-time bias can increase the incidence by a maximum of one multiplied by the mortality rate. Any further increases in incidence might be caused by length-time bias. In table 1, length-time bias might have started to affect 5-year relative survival in the UK, and Austria might be the country where its effect was the highest. In addition to length-time bias, the EUROCARE-4 study only included 1–58% of the population in nine of the countries studied—this has unknown effects on survival estimates. For example, most Italian data came from the north of the country,1 where the incidence of and survival from prostate cancer are higher than in the south.9 Relative survival data are susceptible to biases and limitations of both incidence and mortality indicators.10 Mortality data are usually available for entire countries and not confined to some sub-country areas. Therefore, in the current situation, mortality data remain the most reliable indicator for the international comparison of outcomes in cancer. *Philippe Autier, Mathieu Boniol, Clarisse Héry, Eric Masuyer, Jacques Ferlay International Agency for Research on Cancer, 69372 Lyon Cedex 08, France [email protected] The authors declared no conflicts of interest.
1052
1
2 3 4 5 6
7
8
9
10
Berrino F, De Angelis R, Sant M, et al. Survival for eight major cancers and all cancers combined for European adults diagnosed in 1995–99: results of the EUROCARE-4 study. Lancet Oncol 2007; 8: 773–83. Sackett DL, Haynes RB, Guyatt GH, Tugwell P. Clinical epidemiology: a basic science for clinical medicine, 2nd edn. Boston: Little, Brown, 1991: 153–70. Parkin DM, Whelan SL, Ferlay J, Storm H. Cancer Incidence in Five Continents, Volume I to VIII. Lyon: IARC CancerBase No.7, 2005. World Health Organisation Databank, Geneva, Switzerland. WHO Statistical Information System. http:www-dep.iarc.fr (accessed Nov 9, 2007). Health in the European Union, Report of September 2007. http://ec.europa. eu/health/ph_publication/eb_health_en.pdf (accessed Nov 9, 2007). Fracheboud J, Otto SJ, van Dijk JA, et al. Decreased rates of advanced breast cancer due to mammography screening in the Netherlands. Br J Cancer 2004; 91: 861–67. Helgesen F, Holmberg L, Johansson JE, Bergström R, Adami HO. Trends in prostate cancer survival in Sweden, 1960 through 1988: evidence of increasing diagnosis of nonlethal tumors. J Natl Cancer Inst 1996; 88: 1216–21. Lipsker DM, Hedelin G, Heid E, Grosshans EM, Cribier BJ. Striking increase of thin melanomas contrasts with stable incidence of thick melanomas. Arch Dermatol 1999; 135: 1451–56. Quaglia A, Vercelli M, Puppo A, et al. Prostate cancer in Italy before and during the ‘PSA era’: survival trend and prognostic determinants. Eur J Cancer Prev 2003; 12: 145–52. Doll R, Peto R. The causes of cancer. Appendix C. Oxford: Oxford University Press, 1981.
In a recent issue of The Lancet Oncology, findings from EUROCARE-41 were published, which showed that survival for all cancers combined in the UK was not only below the European average, but also similar to survival in some eastern European countries, where the healthcare budget is less than a third of that in the UK. These findings led to the suggestion that the UK government now faces substantial challenges in improving cancer survival.2 The study also showed that European survival from solid tumours was lower than that reported in the USA.3 The clear implication of these findings is that patients presenting with the same disease will do less well in the UK than in other western countries, a thought that is frightening for UK National Health Service (NHS) patients and demoralising for NHS staff. However, a closer look at the published data suggests that the quality of care offered by the NHS is not the only explanation for poor UK survival, with differences in patient characteristics, disease http://oncology.thelancet.com Vol 8 December 2007